BRAF inhibition in hairy cell leukemia with low-dose vemurafenib

Hairy cell leukemia (HCL) is a mature B-cell lymphoid malignancy presenting with pancytopenia and splenomegaly.^1-3  Standard treatment is based on chemotherapy with purine analogs,^1,2  but eradication of minimal residual disease is rarely achieved by purine analogs alone.⁴ 

Gain-of-function mutations of the BRAF serine/threonine protein kinase (BRAF V600E) have been identified in 95% to 100% of classical HCL.^5,6  The incidence and the presence in all malignant B cells^5,7  suggest that HCL cells critically depend on activated BRAF, providing oncogenic signaling through the MEK–extracellular signal-regulated kinase (ERK) cascade.^5,8  Data from trials in BRAF-mutated melanomas^9,10  have inspired investigators to treat refractory HCL patients with the BRAF inhibitor (BRAFi) vemurafenib, which showed striking clinical activity.^11-15  Dosing of vemurafenib outside clinical trials has been significantly lower than standard melanoma dosing (240 mg twice daily vs 960 mg twice daily).^12,16  Two phase 2 trials (an Italian trial [n = 26] and a US trial [n = 24]) have explored the melanoma dose of 2 × 960 mg and demonstrated efficacy in all treated patients.¹⁷  Despite improvement of blood counts in all patients, complete remission (CR) was achieved by only 35% to 42% of patients. Assessment of minimal residual disease revealed persisting hairy cells, even in patients with CR.¹⁷ 

Maximum tolerated dose traditionally serves as the primary end point for the majority of dose-finding studies, but on-target efficacy often occurs at lower doses. Indeed, dosing of cancer cell–specific drugs may be best assessed by on-target inhibition, response, and safety.^18,19  Dose finding of vemurafenib in BRAF-mutant melanoma was based on dose-limiting toxicities occurring in <30% of patients, but objective responses were observed at the first dose escalation level of 2 × 240 mg.⁹ 

In the current study, we collected a series of HCL patients with long-term follow-up who have been treated at our centers. The data provide conclusive evidence that low doses of vemurafenib are active in HCL. Based on the presence of typical side effects, as well as clinical and experimental evidence for sufficient on-target activity, the data suggest that systematic dose reconsideration is warranted in HCL and potentially other BRAF-mutant cancers, which could significantly reduce cost of care.

We report on 21 patients with classical HCL treated with vemurafenib in 11 different European centers (Heidelberg, n = 6; Innsbruck, n = 4; Nice, n = 2; Munich, n = 2; Cambridge, n = 1; Erfurt, n = 1; Freiburg, n = 1; Lucerne, n = 1; Cologne, n = 1; Leicester, n = 1; and London, n = 1) outside of trials from 2011 to 2014. Five patients have previously been reported and are included with updated follow-up.^12-14,20,21  Clinical data and follow-up information were collected by chart review. Bone marrow slides were centrally reviewed (n = 11, by M.A.). Trephine biopsies were not available for 4 of 15 patients. For these 4 patients, pathology reports were carefully reviewed. Responses were evaluated based on blood counts, bone marrow findings, and peripheral blood hairy cell count using standard criteria.¹  CR was achieved when the platelet counts were >100 000/μL, hemoglobin level was >12 g/dL, neutrophil counts were >1000/μL, spleen size had normalized, and bone marrow biopsy was negative for hairy cells. For the definition of CR, the evaluation of trephine biopsy samples and peripheral blood smears had to be negative for hairy cells, at least based on morphology using nonimmunologic stains.

Median time between start of vemurafenib treatment and bone marrow biopsy was 3 months (range, 1.0-11.1 months). Bone marrow biopsies were performed in the context of recovered blood counts.

To determine spleen size, we used the largest diameter of the spleen, which correlates well with its volumes.²²  Spleen sizes were measured by ultrasound, magnetic resonance imaging, or computed tomography. We used a cutoff for splenomegaly of >13 cm.²²  In 1 patient for whom no ultrasound or other imaging studies were performed, we report reduction of spleen sizes based on physical examination.

Immunohistochemistry for PAX5/phosphorylated ERK (p-ERK), glycophorin/p-ERK, and cyclin D1 were performed according to standard methods (see supplemental Materials and Methods, available on the Blood Web site). Anti-BRAF V600E immunostaining was done as described previously using the monoclonal mouse antibody VE1.²³ 

Further methods are provided in supplemental Materials and Methods.

The presence of the BRAF V600E mutation was demonstrated in all patients by immunohistochemical staining (BRAF V600E mutation–specific antibody) (n = 6) and/or sequencing (n = 15). Patient characteristics are summarized in Table 1. Median age at initiation of vemurafenib was 64 years (range, 45-89 years). Median time from diagnosis to treatment was 8 years (range, 0-31 years). Patients were heavily pretreated (median of 3 prior treatment lines; range, 0-12 lines; n = 19). Indication for treatment was based on presence of cytopenia in all patients (thrombocytopenia <100 000/μL, n = 19/21; hemoglobin <10 g/dL, n = 15/21; or neutrophils <1000/μL, n = 18/21). Based on these criteria, trilineage cytopenias were observed in 14 of 21 patients and bilineage cytopenias in 3 of 21 patients. Two patients were treated with vemurafenib as first-line therapy. Vemurafenib dose was chosen by the authors independently, which allowed us to compare the effect of dose levels. A dose of 240 mg twice daily was used in 17 patients. Twelve patients continued at this dose, whereas doses were escalated in 5 patients (480 mg, n = 1; 720 mg, n = 2; and 960 mg, n = 2; all twice daily). Further patients received 240 mg once daily (n = 1), 480 mg twice daily (n = 2), or 960 mg twice daily (n = 1) without dose modification. No information on why doses were escalated was available. Median duration of treatment was 90 days (range, 56-266 days), with a median cumulative treatment dose of 51 000 mg (range, 27 000-311 000 mg). Discontinuation of therapy with vemurafenib was based on physicians’ choices and on full recovery of blood counts in 20 of 21 patients. One patient developed an acute myeloid lymphoma (AML) subtype M6 with concomitant deteriorating liver function tests, and vemurafenib was subsequently stopped.

Blood counts improved in all patients (Figure 1A; supplemental Figure 1). Platelets recovered first (median time to platelets >100.000/μL, 28 days; range, 10-105 days). Median time to recovery of neutrophils (>1000/μL) was 43 days (range, 9-126 days) and median time to hemoglobin recovery (>12 g/dL) was 55 days (range, 10-181 days) (Figure 1A). There was no detectable difference in recovery of blood count dynamics for patients who received low doses of vemurafenib (≤240 mg, Figure 1A; supplemental Figure 1). In addition, treatment duration as a continuous variable did not have an impact on reconstitution of blood counts (platelets, P = .15; hemoglobin, P = .13; and neutrophils, P = .284).

Hematological response was achieved in 20 of 21 patients (95%). The 1 patient failing to meet these criteria developed AML-M6 after vemurafenib treatment, resulting in inefficient erythropoiesis (see “Side effects during vemurafenib treatment”), whereas platelet and neutrophil counts improved sufficiently to meet response criteria. Based on response assessment of bone marrow trephine biopsy samples, 6 of 15 patients achieved a CR. Although response criteria were otherwise achieved, 6 patients had no trephine biopsy, and formal response could not be determined.

In logistic regression analysis, CR rate was not associated with cumulative vemurafenib dose (Figure 1A; P = .73) or treatment duration (P = .76).

With a median observation time of 17 months, median event-free survival (EFS; start of vemurafenib treatment to retreatment or death) was 17 months (Figure 2). EFS was not influenced by cumulative administered dose of vemurafenib (P = .23; hazard ratio (HR), 0.90; 95% confidence interval (CI), 0.8-1.1; effect size for HR per 10 g vemurafenib) or treatment duration (P = .31; HR, 1.3; 95% CI, 0.6-2.1; effect size for HR per 1 month). Median time to relapse as defined by deterioration of blood counts below remission thresholds was 14 months, and overall survival at 12 months was 88%. Three of 21 patients died (disease progression, n = 1²⁰ ; pneumonia in remission, n = 1; AML, n = 1). Achieving a CR after vemurafenib treatment was associated with better EFS (time from cessation of vemurafenib/response assessment to retreatment or death; P = .04; HR, 0.2; 95% CI, 0.1-0.9).

Nine patients (42%), including the 2 patients treated upfront, were re-treated at relapse (range, 4-17 months). Six patients were reexposed to vemurafenib and responded (Table 2). Two patients received a second course of low-dose vemurafenib, and 4 patients (Heidelberg 03 and 05, Cambridge, and Nice 01) received continuous treatment at either 240 mg once daily or 240 mg twice daily (14, 6, 25 and 9 months from restarting therapy, respectively). Kinetics of response resembled the initial treatment course (Figure 1).

To understand the effect of vemurafenib on downstream BRAF V600E targets, we assessed expression of p-ERK and presence of BRAF V600E–positive cells by immunohistochemistry. Upon vemurafenib treatment, the hairy cell infiltration of the marrow decreased to meet response criteria (>50%, Figure 1B). Using Pax5/p-ERK double staining of trephine biopsy samples, we found complete abrogation of p-ERK in PAX5-positive cells at 240 mg (twice daily) of vemurafenib (Figure 1B). Cyclin D1 expression was undetectable in hairy cells after vemurafenib exposure, suggesting BRAF dependence (patient 01, day 6; patient 02, day 63; and patient 03, day 85).^9,11 

To gain a comprehensive understanding of cytokines involved in HCL, we performed cytokine arrays of serum samples before and upon vemurafenib treatment (supplemental Figure 2). Soluble CD25 (sCD25; interleukin-2 receptor) was downregulated upon vemurafenib treatment, and additional cytokines that decreased with treatment included insulin-like growth factor binding protein 1, soluble tumor necrosis factor receptor type I and II, and B-cell-attracting chemokine 1 (supplemental Figure 2). Brain-derived neurotrophic factor, CCL5/RANTES, epidermal growth factor, and platelet-derived growth factor (supplemental Figure 2) increased upon BRAFi treatment. We repeatedly measured sCD25 serum levels in a subgroup of patients (n = 6). sCD25 levels decreased below the upper normal limit after a median of 36 days (Figure 1C). After cessation of vemurafenib, sCD25 levels rose, exhibiting individual progression dynamics (Figure 1C).

The side-effect profile of vemurafenib included arthralgia (n = 4), which resolved with nonsteroidal anti-inflammatory drugs or low-dose steroids, and mild reversible elevation of liver enzymes (n = 4). An 89-year-old patient developed acute on chronic renal failure during vemurafenib treatment. Phototoxicity (National Cancer Institute Common Toxicity Criteria grade ≤2, n = 4), keratoacanthomas (n = 3), squamous cell papilloma (n = 1), and squamous cell carcinomas (n = 1) occurred (Table 1). Except for 1 patient with keratoacanthoma who received 480 mg of vemurafenib twice daily, all patients with skin tumors received 240 mg twice daily. Progression of non–skin cancer has been reported during vemurafenib treatment,^24,25  which was linked to RAS-mediated paradox ERK activation of wild-type BRAF in the context of vemurafenib. Patient 12 presented with pure red cell AML-M6b accelerated by vemurafenib treatment (Figure 3). While exposed to vemurafenib, the patient had 40% of erythroblasts (glycophorin A⁺, CD36⁺, CD71⁺, CD34⁻; supplemental Figure 3) in the peripheral blood (Figure 3). Lactate dehydrogenase levels and liver enzymes increased on vemurafenib, but after cessation of vemurafenib, both lactate dehydrogenase and transaminases returned to normal levels and erythroblasts disappeared from the peripheral blood (Figure 3). At 1.8 months after discontinuation of vemurafenib, the patient was diagnosed with AML-M6, with massive elevation of liver function tests. The diagnosis was confirmed by immunophenotyping (supplemental Figure 3) and bone marrow histology (Figure 3,). Standard molecular workup revealed a translocation t(20;22)(q13.1;q13). Next-generation sequencing (454 and Ion Torrent Hotspot Panel v2) showed wild-type NRAS and KRAS, but we identified an activating PI3KCA mutation (E545K). We performed deep-sequencing (Ion Torrent, >16 000 reads) using trephine biopsy material taken 44 days before vemurafenib was administered, and could not detect the PI3KCA mutation (E545K) within the sensitivity of the assay used (>0.1%). Vemurafenib can cause paradoxical ERK activation in the context of wild-type BRAF and activated RAS.²⁶  Because phosphatidylinositol 3-kinase (PI3K) has been demonstrated to activate RAS,²⁷  we choose cell line models (MCF-7 and L363) with PI3KCA E545K mutation and exposed the cell lines to BRAFi in vitro. We demonstrated paradoxical ERK activation upon BRAFi treatment similar to a RAS mutant control cell line (NOMO-1). It was unfortunate that viable tissue from the time of vemurafenib treatment was not available for the patient, but immunohistochemistry demonstrated strong p-ERK expression in the glycophorin A⁺ intrasinusoidal erythroblasts in the trephine biopsy material at diagnosis of the AML-M6 (Figure 3).

PI3K mutations have not been associated with AML.²⁸  We therefore analyzed 40 acute erythroid leukemias (37 erythroleukemia M6a and 3 pure erythroid leukemia M6b) and found no additional PI3K mutations, suggesting that these mutations are not recurrent at higher frequencies in AML-M6.

We describe the outcome of vemurafenib treatment in 21 patients with HCL. Our findings indicate that a short course of BRAF inhibition with a low-dose of vemurafenib can effectively inhibit MEK/ERK signaling in vivo, reduce HCL load, and induce CRs in HCL patients. These data confirm the critical dependence on active BRAF signaling in HCL.

Despite rapid improvement of blood counts, CR was achieved in only 40% of patients, which is in line with the CR rates reported in the prospective phase 2 trials.¹⁷  Although mechanisms underlying disease persistence are currently unclear, our data suggest that insufficient dosing is unlikely to be the cause for heterogeneous response. Immunohistochemistry showed completely abrogated ERK activation, indicating on-target efficacy of low-dose vemurafenib. In line with these observations, sCD25 (sIL2R-α) levels were rapidly reduced with vemurafenib, including patients with persisting hairy cell infiltration. In melanoma, combination treatments of BRAFi and MEKi have been demonstrated to improve response rates and survival.²⁹  Whether response rates can also be improved in HCL is currently being tested in clinical trials (eg, NCT02034110).

Although we observed no progressions on drug, the median time to retreatment after cessation of vemurafenib was 14 months. Residual cells appear to reenter proliferation and give rise to disease progression when treatment is stopped. To develop more effective treatment schedules for BRAF inhibition or combination treatment, it would be critical to understand why HCL cells persist in the presence of inhibited MEK/ERK signaling. Potential explanations include failure to execute cell death upon oncogene inhibition (eg, with vemurafenib).¹¹  BRAF V600E has been shown to suppress the cell cycle regulator p27,³⁰  which is recurrently mutated in HCL,³¹  suggesting that cell cycle control or oncogene-induced senescence might be alternative fail-safe mechanisms. Failure to clear disease could also be based on alternative survival pathways; for example, microenvironmental signals or B-cell receptor signaling.³² 

The vemurafenib schedule (960 mg twice daily) was derived from a phase 1 trial in melanoma and was based on the predefined incidence of dose-limiting toxicities. Concerns have been raised about whether dose-finding strategies based on dose-limiting toxicities as established for chemotherapy drugs can be adopted for targeted drugs.^18,19  Two prospective studies have used the standard dose of vemurafenib (960 mg twice daily) and report response rates and relapse-free survival rates strikingly similar to those of the current study.¹⁷  In the prospective studies, dose reduction was necessary in >50% of patients (Italian trial, 14/26; US trial, 17/28) and no difference in outcome of patients receiving the full or the reduced dose could be demonstrated.

Long-term treatment of cancer with BRAFi is challenged by secondary resistance formation^33-35  and development of secondary tumors based on paradoxical ERK activation in cells with wild-type BRAF. Skin cancers have been reported at frequencies of 15% to 30% in BRAF-mutant melanoma patients receiving BRAFi,³⁶  and we observed a comparable frequency (24%). Low doses may therefore cause on-target, but also comparable, rates of side effects. In tumors and normal cells with wild-type RAF, BRAFi paradoxically stimulates ERK signaling in a RAS-dependent manner.^26,37  In addition, case reports describe patients who experienced progression of preexisting RAS-mutated (chronic myelomonocytic leukemia) or RAS pathway–driven malignancies (eg, pancreatic cancer³⁸ ) during BRAFi treatment.^25,38  BRAFi-mediated paradoxical ERK activation was shown to be B-cell-receptor dependent in chronic lymphocytic leukemia cells.²⁴ 

We provide further in vivo insight on the risk of BRAFi-driven malignancies and report 1 patient who developed an AML-M6b (subtype pure red cell AML) during vemurafenib treatment. In line with the concept of enhanced paradoxical ERK activation in the context of activated RAS, we identified a PI3K (E545K) mutation in the emerging AML clone, which is known to activate RAS.²⁷  Although we cannot directly demonstrate p-ERK activation in erythroblasts during vemurafenib treatment, our results show ERK activation in AML cells. The biphasic course and the regression after vemurafenib cessation suggest that vemurafenib may have contributed to AML-M6 development. These data support the use of MEKi also in HCL, in order to reduce the risk of paradoxical ERK activation.³⁹  The majority of reports on BRAFi-induced secondary malignancies describe preexisting malignant or premalignant clones, which experience accelerated clonal expansion upon BRAFi exposure.^25,38  Although we were not able to detect an AML clone before vemurafenib treatment based on very deep sequencing of the PI3K mutation, we cannot prove the preexistence of premalignant precursors. With 5 prior lines of chemotherapy, a contribution of DNA-damaging agents to the formation of AML cannot be excluded.

We report stable long-term remissions on low-dose vemurafenib, but continuous treatment involves the risk of resistance formation and secondary malignancies. This risk might be reduced with altered on-off dosing schedules Experimental models of BRAF inhibition have shown a reduction in resistance formation when BRAFi are applied in on-off schedules.⁴⁰ 

Despite the retrospective nature and the number of patients studied, our analysis provides starting points for the reevaluation of traditional approaches to dose finding for targeted cancer drugs. Although not a clinical trial, the individual dosing regimens allowed us to systematically assess the impact of dosing on clinical and pharmacodynamic end points. Major innovations in targeted treatment^41,42  have stemmed from clinical observations outside of trials. The results of the current work support detailed dose-finding studies assessing molecular-target inhibition and side-effect profile when using targeted anticancer drugs. The data also suggest that optimal dosing schedules for vemurafenib may need to be reassessed in clinical trials, which could have major pharmacoeconomic implications.

Future trials exploring precision medicine should accommodate flexible end points integrating biomarker assessment of on-target effects together with traditional response assessment.

The authors thank T. Uhrig for technical assistance with immunohistochemical stainings, and B. Falini and E. Tiacci (Institute of Hematology, University of Perugia, Perugia, Italy) for the double immunohistochemical staining for glycophorin and PAX5 and the review of bone marrow slides.

This work was supported by a Heidelberg Research Centre for Molecular Medicine grant (S.D.); the Deutsche Krebshilfe (Mildred Scheel Professorship [T.Z.]); the Hairy Cell Foundation; Molecular Medicine Partnership Unit (European Molecular Biology Laboratory and University Hospital Heidelberg); and a Leicester Experimental Cancer Medicine Centre grant (C325/A15575 Cancer Research UK/Department of Health, United Kingdom).

Contribution: S.D. and T.Z. designed and performed the research, provided the samples, and wrote the paper; A.P., G.A.F., J.H., A.J., T.W., T.E., T.B., R.Z., W.W., C.v.K., A.D.H., and M. Seiffert performed the research and edited the paper; E.E., V.E., M.A., and M. Steurer performed the research and wrote the paper; and A.D.H., C.-M.W., F.P., M. Hermann, M. Herold, C.D., T.H., M. Hallek, X.L., M.J.S.D., and A.G. provided the samples and edited the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Thorsten Zenz, Department of Translational Oncology, National Center for Tumor Diseases and German Cancer Research Center, Im Neuenheimer Feld 460, 69120 Heidelberg, Germany; e-mail: thorsten.zenz@nct-heidelberg.de.